In the past the refereeing had two aims:
The reader of a scientific published article wanted a guarantee,
that the article has been screened by some referees.
By picking the journal and knowing its past reputation, he had a feeling
by his or other's experience what level of srcutiny to expect from the
Publisher.
The author wanted to reach out to his readers and by getting his article
into the suitable journal got this label, his aimed at readers use as a
filter.
In addition the author needs the thus gained reputation for his applications
for positions, etc. The hiring committees as well used the reputation of
puthor's picked journals to help judging.
Thus the peer refereeing system was basically attached to the long standing
and developed history and reputation of the scientific international journals.
Without having a better and worldwide established system to cope with these
job-related tasks as well, one has here to keep in parallel to new
expermental ways the traditional way, added by a smooth extension of the
journals being offered in parallel electronically and archived that way.
However, experimentally, and in parallel, the Physics Societies could think of setting up a Peer system
by themselves, -as APS has anyway due to its journals-, which could
be quite differently organized:
Hiring commitees of a department, say, could ask the Society for a scientific
referee report for the applicants. It would delegate this to its respective division (at present for APS its PR, PRL journal refereeing division), picking
the actual referees by the well established patterns, who would
then greatly ease their work by having full access to the work of the candidate by the internet accessed archive.
The innovative part is more subtle: authors could decide to live with
their papers as ep-preprints, with no delay in distribution by refereeing.
But they could ask the referee system for screening specific articles,
in order to eventually address a reader having turned on his filters
to get only refereed articles.
The screening and the distribution of science would thus be decoupled,
and not more refereeing been done as necessary.